Bengaluru-based Sarvam AI has released Bulbul v1, a new text-to-speech AI model that supports 11 Indian languages, including Hindi, Marathi, Punjabi, Oriya, Tamil, Bengali, Telugu, Kannada, Malayalam, and Gujarati. This model features voices with authentic and non-robotic accents.

Other than creating their own voices, users can select from six unique personalities designed for different work needs across different industries. This comes with the added benefits of faster response times and lower costs compared to alternatives such as Eleven Labs.

According to their recent blog post, Sarvam used a high-quality and diverse audio collection to train their AI model.

This dataset has audio from several speakers in many languages. It intentionally includes code-mixed language, names, short forms, and voices that sound both formal and casual for various uses.

In a significant move last week, the government selected Sarvam out of 67 applications to develop India’s first homegrown AI large language model. 

Recognising the ambitious task of building a foundational AI model from scratch, the government is backing Sarvam AI with the significant computing power required.

Source: 

https://www.thehindu.com/sci-tech/technology/sarvam-ai-launches-ai-text-to-speech-model-with-support-for-11-indian-languages/article69548709.ece